Traveler safety early warning method and system based on real-time location data management

Abstract

The application discloses a kind of travel personnel safety early warning method and system based on real-time position data management.The method is by constructing travel personnel position collection platform, obtains the GPS positioning data and actual position data of travel personnel, forms positioning accuracy characteristic map and data transmission performance characteristic map;Gaussian process regression is used to fuse the atlas, generate joint feature vector, determine the collection accuracy offset value of different positions;Further combined with decision tree model to establish personnel position offset prediction model, predict the actual position of travel personnel;Based on the prediction result, the positioning data of the collection platform is corrected to obtain higher-precision position data;Finally, according to the correction data, real-time safety warning is carried out on travel personnel.This method can effectively overcome the error caused by the fluctuation of positioning accuracy and unstable transmission performance in the complex environment of scenic spots, improve the accuracy and reliability of travel personnel positioning.

Description

Technical Field

[0001] This invention relates to the field of personnel safety early warning technology, and in particular to a method and system for travel personnel safety early warning based on real-time location data management. Background Technology

[0002] With the rapid development of the tourism industry, the safety of tourists within scenic areas has gradually become a key concern for managers and the public. To achieve safe management and risk warning for tourists, personnel monitoring technology based on real-time location data has been widely applied. Currently, commonly used methods for locating tourists mainly rely on GPS positioning technology from mobile terminals and pre-installed auxiliary positioning systems within scenic areas. However, in practical applications, tourist attractions are often characterized by complex environments, such as tall buildings, dense vegetation, and terrain obstructions. These factors can lead to GPS signal attenuation or multipath effects, resulting in significant fluctuations in positioning accuracy and making it difficult to guarantee the accuracy of personnel location information.

[0003] Furthermore, the spatial and temporal distribution of tourists within scenic areas typically varies significantly. When visitor traffic is concentrated in specific areas, location data collection and transmission are prone to congestion, leading to increased data transmission latency or even data loss, further impacting the reliability of real-time monitoring and safety alerts. The cumulative errors caused by fluctuations in positioning accuracy and unstable data transmission performance not only reduce the effectiveness of tourist location monitoring but may also lead to delayed or false alarms in safety alerts, failing to meet the scenic area's needs for efficient and safe management.

[0004] Therefore, how to effectively overcome the problems of fluctuating positioning accuracy and unstable positioning data transmission in the complex environment of scenic spots, improve the accuracy and real-time performance of traveler location monitoring, and thus achieve scientific safety early warning for travelers has become an urgent technical problem to be solved. Summary of the Invention

[0005] To address at least one of the aforementioned technical problems, this invention proposes a traveler safety early warning method and system based on real-time location data management.

[0006] The first aspect of this invention provides a traveler safety early warning method based on real-time location data management, comprising: Construct a location data collection platform for tourists at the target scenic spot, acquire GPS positioning data and actual location data of tourists within the target scenic spot from the location data collection platform, determine the positioning accuracy characteristics and positioning data transmission performance characteristics of different locations at the target scenic spot, and construct a positioning accuracy feature map and a positioning data transmission performance feature map; The positioning accuracy feature map and the positioning data transmission performance feature map are fused based on Gaussian process regression to construct a joint feature vector. The positioning data acquisition accuracy offset value of the location acquisition platform at different locations is determined based on the joint feature vector, including the offset direction and offset distance value. Based on the decision tree model and the location acquisition platform's accuracy offset values ​​for different locations, a personnel location offset prediction model is constructed. Based on the personnel location offset prediction model, the location of travelers on the location acquisition platform is predicted, and the predicted actual location data is output. The traveler's location on the location acquisition platform is corrected based on the predicted actual location data to obtain corrected location data; Safety warnings are issued to travelers at the target attractions based on the corrected location data.

[0007] In this solution, the step of constructing a tourist location collection platform for the target scenic spot involves acquiring GPS positioning data and actual location data of tourists within the target scenic spot from the location collection platform, determining the positioning accuracy characteristics and positioning data transmission performance characteristics of different locations within the target scenic spot, and constructing a positioning accuracy feature map and a positioning data transmission performance feature map. Specifically: A location data collection platform for tourists at the target scenic spot is constructed. The location data collection permissions of the mobile terminals of tourists entering the target scenic spot are obtained. The location data collection permissions are connected to the location data collection platform to collect the GPS positioning data of tourists in real time. At the same time, the actual location data of tourists are obtained based on the pre-installed visual positioning system in the target scenic spot. The actual location data distribution density of each area within the target scenic spot is calculated based on the kernel density estimation algorithm within a first preset time period. The target scenic spot is then divided into several location analysis areas based on the actual location data distribution density. Perform spatiotemporal alignment of GPS positioning data and actual location data at the same timestamp to obtain a set of location data pairs. Within each positioning analysis area, calculate the mean Euclidean distance between GPS positioning data and actual location data in each location data pair as the positioning accuracy feature of that area. The generation timestamp of GPS positioning data and the reception timestamp of GPS positioning data of the location acquisition platform are obtained for each positioning analysis area within a second preset time period. Based on the generation timestamp and the reception timestamp, the data transmission delay variation data within the second preset time period is determined. Based on the real-time location data, determine the pedestrian flow change data of each location analysis area within a second preset time period, perform correlation analysis between the pedestrian flow change data and the data transmission delay change data, establish a pedestrian flow-transmission delay correlation model, and obtain the location data transmission performance characteristics of each location analysis area under different pedestrian flows. Based on the Kriging interpolation method, the positioning accuracy features and positioning data transmission performance features are interpolated within the positioning analysis area of ​​the target scenic spot, and mapped to the position coordinates of each positioning analysis area of ​​the target scenic spot to generate the positioning accuracy feature map and the positioning data transmission performance feature map of the target scenic spot.

[0008] In this scheme, the positioning accuracy feature map and the positioning data transmission performance feature map are fused based on Gaussian process regression to construct a joint feature vector. Based on this joint feature vector, the positioning data acquisition accuracy offset value of the location acquisition platform for different locations is determined, including the offset direction and offset distance value. Specifically: A Gaussian process regression algorithm is introduced, and the position coordinates of each positioning analysis region in the positioning accuracy feature map and the positioning data transmission performance feature map are used as the input space of the Gaussian process regression. Based on the positioning accuracy feature map and the positioning data transmission performance feature map, the positioning accuracy feature value and the positioning data transmission performance feature value of each positioning analysis region are extracted as observation values ​​to construct a training dataset. The covariance function of Gaussian process regression is initialized as a radial basis function. Initial hyperparameters, including length scale and signal variance, are set. The hyperparameters of Gaussian process regression are optimized by the maximum likelihood estimation method to obtain the optimized covariance function. The joint covariance matrix between the positioning accuracy feature value and the positioning data transmission performance feature value is calculated based on the optimized covariance function. The Cholesky decomposition of the joint covariance matrix is ​​used to obtain the lower triangular matrix. The weight coefficients of the Gaussian process regression are then solved based on the lower triangular matrix. Input the location coordinates of each location analysis area into the trained Gaussian process regression model, and output the fused prediction value of the location accuracy feature and the location data transmission performance feature. Construct a joint feature vector for each location analysis area based on the fused prediction value. The feature gradient change between adjacent positioning analysis areas is calculated based on the joint feature vector. The positioning data acquisition accuracy offset direction is solved by the gradient descent method. The gradient change rate of the joint feature vector in the offset direction is calculated. The gradient change rate is multiplied by the preset offset distance conversion coefficient to obtain the offset distance value. The positioning data acquisition accuracy offset value is constructed by the positioning data acquisition accuracy offset direction and the offset distance value.

[0009] In this solution, a personnel location offset prediction model is constructed based on a decision tree model and the location acquisition platform's positioning data acquisition accuracy offset values ​​for different locations. Based on this personnel location offset prediction model, the location of travelers on the location acquisition platform is predicted, and the predicted actual location data is output. Specifically: The location acquisition platform collects positioning data at different locations according to a preset data volume ratio to construct the training data matrix and validation set data of the decision tree model. The positioning data collection accuracy offset value of each positioning analysis area is used as the input feature, and the Euclidean distance between the GPS positioning data and the actual location data of the corresponding area is used as the output label. The Gini coefficient is used as the splitting criterion for decision tree nodes. The training data matrix is ​​recursively divided into features. At each node, the feature of the positioning data acquisition accuracy offset value that makes the Gini coefficient decrease the most is selected for splitting until the preset tree depth or node sample number threshold is reached, and the initial decision tree model is generated. The initial decision tree model is optimized by post-pruning algorithm. The change in validation set error after each non-leaf node is replaced with a leaf node is calculated. When the error change is less than a preset threshold, pruning operation is performed to obtain the personnel position offset prediction model. The system acquires real-time GPS positioning data of travelers collected by the location acquisition platform, extracts the positioning data acquisition accuracy offset value feature of the positioning analysis area where the GPS positioning data is located, inputs the positioning data acquisition accuracy offset value feature into the trained personnel location offset prediction model, matches the feature value layer by layer through the splitting rules of the decision tree, and finally reaches the leaf node to obtain the corresponding predicted offset distance and predicted offset direction. The predicted actual location data of travelers is obtained by vector superposition calculation based on the coordinate position of real-time GPS positioning data, the predicted offset distance, and the predicted offset direction.

[0010] In this solution, the step of correcting the traveler's location on the location acquisition platform based on the predicted actual location data to obtain corrected location data specifically involves: The predicted actual location data is compared with the GPS location data of travelers collected in the location acquisition platform to determine the deviation between the GPS location data collected by the location acquisition platform and the predicted actual location data. If the deviation value is greater than the preset value, the GPS location data collected by the location acquisition platform is replaced according to the predicted actual location data to obtain the corrected location data of the traveler in the location acquisition platform.

[0011] In this solution, the step of issuing safety warnings to travelers at the target scenic spot based on the corrected location data specifically includes: Construct a set of electronic fence area boundary coordinates based on the safe location boundary of the target attraction, and calculate the minimum Euclidean distance between the traveler's location and the electronic fence area boundary coordinates based on the corrected location data; When the minimum Euclidean distance is less than a preset distance threshold, the traveler's movement trajectory data is acquired, and the movement trend is predicted based on the movement trajectory data using a time series analysis algorithm to obtain the traveler's predicted movement trajectory. The predicted time for a traveler to reach the electronic fence boundary area is calculated based on the predicted movement trajectory. When the predicted time is less than the time warning threshold, the traveler is marked as a location risk person, a location warning information of the location risk person is generated and sent to the staff device, and a warning call alarm is issued to the location risk person's mobile terminal. When the minimum Euclidean distance is not less than a preset distance threshold, the traveler is marked as a person in a safe location, and the flow of people in the location analysis area where the person in the safe location is located is monitored in real time. The flow of people is compared with the preset flow of people threshold. When the change in pedestrian traffic exceeds the pedestrian traffic threshold, the performance characteristics of the location data transmission in that area are extracted. The location data transmission delay is predicted by combining the current change in pedestrian traffic data. Based on the predicted location data transmission delay, the frequency of location data collection by the location acquisition platform for people in safe locations is reduced, thus forming a safety early warning strategy for tourists at the target scenic spot.

[0012] A second aspect of the present invention also provides a traveler safety early warning system based on real-time location data management. The system includes a memory and a processor. The memory includes a traveler safety early warning method program based on real-time location data management. When executed by the processor, the traveler safety early warning method program based on real-time location data management performs the following steps: Construct a location data collection platform for tourists at the target scenic spot, acquire GPS positioning data and actual location data of tourists within the target scenic spot from the location data collection platform, determine the positioning accuracy characteristics and positioning data transmission performance characteristics of different locations at the target scenic spot, and construct a positioning accuracy feature map and a positioning data transmission performance feature map; The positioning accuracy feature map and the positioning data transmission performance feature map are fused based on Gaussian process regression to construct a joint feature vector. The positioning data acquisition accuracy offset value of the location acquisition platform at different locations is determined based on the joint feature vector, including the offset direction and offset distance value. Based on the decision tree model and the location acquisition platform's accuracy offset values ​​for different locations, a personnel location offset prediction model is constructed. Based on the personnel location offset prediction model, the location of travelers on the location acquisition platform is predicted, and the predicted actual location data is output. The traveler's location on the location acquisition platform is corrected based on the predicted actual location data to obtain corrected location data; Safety warnings are issued to travelers at the target attractions based on the corrected location data.

[0013] This invention discloses a method and system for traveler safety early warning based on real-time location data management. The method constructs a traveler location acquisition platform to obtain GPS positioning data and actual location data of travelers, forming a positioning accuracy feature map and a data transmission performance feature map. Gaussian process regression is used to fuse the maps, generating a joint feature vector to determine the acquisition accuracy offset value at different locations. Furthermore, a personnel location offset prediction model is established using a decision tree model to predict the actual location of travelers. Based on the prediction results, the positioning data of the acquisition platform is corrected to obtain higher accuracy location data. Finally, real-time safety early warnings are issued to travelers based on the corrected data. This method can effectively overcome the errors caused by fluctuations in positioning accuracy and unstable transmission performance in complex scenic environments, improving the accuracy and reliability of traveler positioning. Attached Figure Description

[0014] Figure 1 A flowchart of a traveler safety early warning method based on real-time location data management according to the present invention is shown; Figure 2 The flowchart illustrating the present invention for predicting the location of travelers on a location acquisition platform is shown. Figure 3 A flowchart illustrating the process of obtaining corrected position data according to the present invention is shown; Figure 4 A block diagram of a traveler safety early warning system based on real-time location data management according to the present invention is shown. Detailed Implementation

[0015] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0016] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0017] Figure 1 A flowchart of a traveler safety early warning method based on real-time location data management according to the present invention is shown.

[0018] like Figure 1 As shown, the first aspect of the present invention provides a traveler safety early warning method based on real-time location data management, comprising: S102, Construct a location acquisition platform for tourists at the target scenic spot, acquire GPS positioning data and actual location data of tourists within the target scenic spot from the location acquisition platform, determine the positioning accuracy characteristics and positioning data transmission performance characteristics of different locations at the target scenic spot, and construct a positioning accuracy feature map and a positioning data transmission performance feature map; S104, Based on Gaussian process regression, the positioning accuracy feature map and the positioning data transmission performance feature map are fused to construct a joint feature vector. Based on the joint feature vector, the positioning data acquisition accuracy offset value of the position acquisition platform at different locations is determined, including the offset direction and offset distance value. S106, Based on the decision tree model and the location acquisition platform's accuracy offset value for different locations, a personnel location offset prediction model is constructed. Based on the personnel location offset prediction model, the location prediction of the travelers on the location acquisition platform is performed, and the predicted actual location data is output. S108, Correct the traveler's location on the location acquisition platform based on the predicted actual location data to obtain corrected location data; S110, Based on the corrected location data, issue a safety warning to travelers at the target attraction.

[0019] It should be noted that in complex scenic environments, factors such as tall buildings, vegetation obstruction, and concentrated crowds can easily lead to signal attenuation and multipath effects in single GPS positioning, resulting in insufficient positioning accuracy. Furthermore, data transmission may be delayed due to network congestion, further complicating the real-time location tracking of travelers. By constructing a traveler location acquisition platform, which integrates collected GPS positioning data with actual location data, positioning accuracy feature maps and data transmission performance feature maps are obtained. Gaussian process regression is used to fuse these feature maps and generate a joint feature vector, enabling dynamic calculation of acquisition accuracy offset values ​​for different locations. A decision tree model is introduced to construct a personnel location offset prediction model, predicting real-time positioning errors based on historical offset characteristics, thus improving the reliability of location prediction in complex scenarios. Subsequently, by correcting the acquisition platform data with predicted actual location data, the impact of accumulated errors on monitoring results is effectively reduced, achieving high-precision real-time personnel location output. Finally, safety warnings are implemented based on the corrected location data, identifying risks and issuing alerts in advance when travelers approach dangerous boundaries or densely populated areas, avoiding delayed warnings or high false alarm rates, and achieving safety protection for travelers within the scenic area.

[0020] According to an embodiment of the present invention, the step of constructing a traveler location acquisition platform for a target scenic spot involves acquiring GPS positioning data and actual location data of travelers within the target scenic spot from the location acquisition platform, determining the positioning accuracy characteristics and positioning data transmission performance characteristics of different locations within the target scenic spot, and constructing a positioning accuracy feature map and a positioning data transmission performance feature map, specifically as follows: A location data collection platform for tourists at the target scenic spot is constructed. The location data collection permissions of the mobile terminals of tourists entering the target scenic spot are obtained. The location data collection permissions are connected to the location data collection platform to collect the GPS positioning data of tourists in real time. At the same time, the actual location data of tourists are obtained based on the pre-installed visual positioning system in the target scenic spot. The actual location data distribution density of each area within the target scenic spot is calculated based on the kernel density estimation algorithm within a first preset time period. The target scenic spot is then divided into several location analysis areas based on the actual location data distribution density. Perform spatiotemporal alignment of GPS positioning data and actual location data at the same timestamp to obtain a set of location data pairs. Within each positioning analysis area, calculate the mean Euclidean distance between GPS positioning data and actual location data in each location data pair as the positioning accuracy feature of that area. The generation timestamp of GPS positioning data and the reception timestamp of GPS positioning data of the location acquisition platform are obtained for each positioning analysis area within a second preset time period. Based on the generation timestamp and the reception timestamp, the data transmission delay variation data within the second preset time period is determined. Based on the real-time location data, determine the pedestrian flow change data of each location analysis area within a second preset time period, perform correlation analysis between the pedestrian flow change data and the data transmission delay change data, establish a pedestrian flow-transmission delay correlation model, and obtain the location data transmission performance characteristics of each location analysis area under different pedestrian flows. Based on the Kriging interpolation method, the positioning accuracy features and positioning data transmission performance features are interpolated within the positioning analysis area of ​​the target scenic spot, and mapped to the position coordinates of each positioning analysis area of ​​the target scenic spot to generate the positioning accuracy feature map and the positioning data transmission performance feature map of the target scenic spot.

[0021] It should be noted that the traveler location collection platform obtains location permissions from travelers' mobile terminals and transmits their GPS location information to the platform via the mobile terminal network for display. Therefore, the GPS positioning accuracy and the varying pedestrian traffic at different locations within the scenic area will affect the real-time location data received by the platform. The platform uses a kernel density estimation algorithm to analyze the distribution of actual location data, dividing the scenic area into multiple location analysis regions. It then uses a spatiotemporal alignment method to pair GPS data with actual location data at the same timestamp, thereby calculating the positioning accuracy characteristics of each region. Simultaneously, it combines the time difference between GPS data generation and reception within different time periods to obtain the dynamic changes in transmission delay, and analyzes its correlation with pedestrian traffic data to establish a model linking pedestrian traffic and transmission delay, thus obtaining the positioning data transmission performance characteristics under different pedestrian densities. Finally, the platform uses Kriging interpolation to spatially map these characteristics across the entire scenic area, constructing a positioning accuracy feature map and a positioning data transmission performance feature map. This process quantifies and visualizes the differences in positioning accuracy and transmission performance across different areas within the scenic area, effectively revealing the patterns of positioning accuracy fluctuations and data transmission instability in complex environments. The pre-set visual positioning system consists of an array of cameras distributed throughout the scenic spot.

[0022] According to an embodiment of the present invention, the method of fusing the positioning accuracy feature map and the positioning data transmission performance feature map based on Gaussian process regression to construct a joint feature vector, and determining the positioning data acquisition accuracy offset value of the location acquisition platform at different locations based on the joint feature vector, including the offset direction and offset distance value, specifically: A Gaussian process regression algorithm is introduced, and the position coordinates of each positioning analysis region in the positioning accuracy feature map and the positioning data transmission performance feature map are used as the input space of the Gaussian process regression. Based on the positioning accuracy feature map and the positioning data transmission performance feature map, the positioning accuracy feature value and the positioning data transmission performance feature value of each positioning analysis region are extracted as observation values ​​to construct a training dataset. The covariance function of Gaussian process regression is initialized as a radial basis function. Initial hyperparameters, including length scale and signal variance, are set. The hyperparameters of Gaussian process regression are optimized by the maximum likelihood estimation method to obtain the optimized covariance function. The joint covariance matrix between the positioning accuracy feature value and the positioning data transmission performance feature value is calculated based on the optimized covariance function. The Cholesky decomposition of the joint covariance matrix is ​​used to obtain the lower triangular matrix. The weight coefficients of the Gaussian process regression are then solved based on the lower triangular matrix. Input the location coordinates of each location analysis area into the trained Gaussian process regression model, and output the fused prediction value of the location accuracy feature and the location data transmission performance feature. Construct a joint feature vector for each location analysis area based on the fused prediction value. It should be noted that the fused prediction value refers to a comprehensive positioning error assessment index for the location acquisition platform to obtain the location of travelers. This prediction value reflects the location acquisition error of a specific location within the target scenic spot under the combined effect of GPS positioning accuracy and positioning data transmission performance. Specifically, the positioning accuracy feature represents the average deviation between the GPS positioning data and the actual location in the area, while the positioning data transmission performance feature characterizes the degree to which the delay from GPS data generation to reception in the area is affected by pedestrian traffic. Gaussian process regression establishes a nonlinear correlation between these two features in spatial location and uses a covariance function to characterize their joint distribution characteristics. The final output fused prediction value is a weighted result that comprehensively considers positioning deviation and transmission delay. The positioning accuracy deviation coefficient quantifies the inherent error of GPS signal caused by environmental interference, while pedestrian traffic assesses the weakening effect of data transmission delay on positioning real-time performance. The product of the two ensures that the fused prediction value can reflect both static positioning error and dynamic transmission loss. The joint feature vector is the overall influence of positioning data transmission performance and GPS positioning accuracy at different locations and different pedestrian traffic on the location acquisition platform's acquisition of personnel positioning data.

[0023] The feature gradient change between adjacent positioning analysis areas is calculated based on the joint feature vector. The positioning data acquisition accuracy offset direction is solved by the gradient descent method. The gradient change rate of the joint feature vector in the offset direction is calculated. The gradient change rate is multiplied by the preset offset distance conversion coefficient to obtain the offset distance value. The positioning data acquisition accuracy offset value is constructed by the positioning data acquisition accuracy offset direction and the offset distance value.

[0024] It should be noted that the joint feature vector integrates positioning accuracy features and positioning data transmission performance features, and can characterize the trend of positioning data quality changes in different areas of the target scenic spot. Since positioning errors usually exhibit spatial continuity, the differences in joint feature vectors between adjacent areas reflect the potential directional patterns of positioning accuracy offset. By calculating the gradient change of the joint feature vectors of adjacent positioning analysis areas, the rate and direction of change of positioning data acquisition errors in space can be quantified. The gradient descent method, through iterative search along the negative gradient direction, can converge to the direction of minimizing error, thereby solving for the optimal offset direction of positioning data acquisition accuracy. Since the offset distance of positioning error is positively correlated with the gradient change rate of the joint feature vector, the larger the gradient change rate, the more drastic the spatial change of positioning error, and the larger the offset distance. A regression model between the gradient change rate and the actual positioning error is established by training with historical data, and an offset distance conversion coefficient is fitted so that the gradient change rate multiplied by this coefficient can accurately reflect the actual offset distance. After determining the offset direction, the gradient change rate of the joint feature vector is calculated along this direction, and a linear mapping is performed in combination with the offset distance conversion coefficient to finally output the offset distance value. The positioning data acquisition accuracy offset value is a positioning position offset value that includes direction representation. It is a comprehensive position deviation value that reflects the accuracy of the personnel positioning data acquired by the positioning acquisition platform in different locations, considering both the positioning data transmission performance and GPS positioning accuracy.

[0025] Figure 2 The flowchart illustrating the present invention for predicting the location of travelers on a location acquisition platform is shown.

[0026] According to an embodiment of the present invention, the step of constructing a personnel location offset prediction model based on a decision tree model and the location acquisition platform's positioning data acquisition accuracy offset values ​​for different locations, and using the personnel location offset prediction model to predict the location of travelers on the location acquisition platform, and outputting predicted actual location data, specifically involves: S202, the location acquisition platform collects positioning data accuracy offset values ​​at different locations according to a preset data volume ratio to construct a training data matrix and validation set data for a decision tree model. The positioning data acquisition accuracy offset value of each positioning analysis area is used as the input feature, and the Euclidean distance between the GPS positioning data and the actual location data of the corresponding area is used as the output label. S204 uses the Gini coefficient as the splitting criterion for decision tree nodes, recursively splits the training data matrix by feature division, selects the positioning data acquisition accuracy offset value feature that causes the Gini coefficient to decrease the most at each node for splitting, until the preset tree depth or node sample number threshold is reached, and generates the initial decision tree model. S206. The initial decision tree model is optimized by post-pruning algorithm. The error change of the validation set after each non-leaf node is replaced with a leaf node is calculated. When the error change is less than a preset threshold, pruning operation is performed to obtain the personnel position offset prediction model. S208: Obtain real-time GPS positioning data of travelers collected by the location acquisition platform, extract the positioning data acquisition accuracy offset value feature of the positioning analysis area where the GPS positioning data is located, input the positioning data acquisition accuracy offset value feature into the trained personnel position offset prediction model, match the feature value layer by layer through the splitting rules of the decision tree, and finally reach the leaf node to obtain the corresponding predicted offset distance and predicted offset direction. S210: Based on the coordinates of the real-time GPS positioning data, the predicted offset distance, and the predicted offset direction, vector superposition calculation is performed to obtain the predicted actual location data of the traveler.

[0027] It should be noted that, due to the accuracy offset of GPS positioning data at different locations, and the learnable mapping relationship between these offsets and the actual location, a decision tree model can be constructed by using the accuracy offset values ​​of positioning data collection in each positioning analysis area as input features and the Euclidean distance between the corresponding area's GPS positioning data and the actual location data as output labels. This allows for recursive partitioning of the feature space, selecting the offset value feature that best reduces prediction errors for node splitting, and optimizing the model through post-pruning to obtain an accurate personnel location offset prediction model. This model is then used to predict the GPS data of travelers collected in real-time by the location acquisition platform. The accuracy offset values ​​of the location data collection in the area are input into the model, and the leaf nodes are matched using the decision tree's splitting rules to obtain the predicted offset direction and distance. These are then vector-superimposed with the real-time GPS coordinates to obtain the predicted actual location data of the travelers, thus effectively correcting positioning errors and improving the real-time accuracy of travelers' locations. The data training matrix represents locations in rows and pedestrian traffic in columns, with the matrix elements being the accuracy offset values ​​of the positioning data collection.

[0028] Figure 3 A flowchart illustrating the process of obtaining corrected position data according to the present invention is shown.

[0029] According to an embodiment of the present invention, the step of correcting the traveler's location on the location acquisition platform based on the predicted actual location data to obtain corrected location data specifically involves: S302, compare the predicted actual location data with the GPS location data of travelers collected in the location acquisition platform to determine the deviation value between the GPS location data collected by the location acquisition platform and the predicted actual location data. S304, if the deviation value is greater than the preset value, the GPS location data collected by the location acquisition platform is replaced according to the predicted actual location data to obtain the corrected location data of the traveler in the location acquisition platform.

[0030] It should be noted that by comparing the predicted actual location data with the collected GPS location data, calculating the deviation between the two, and replacing the original GPS data with the predicted actual location data when the deviation exceeds a preset threshold, the location of travelers can be corrected. This can significantly improve the accuracy and reliability of location data and eliminate the positioning offset problem caused by GPS errors and positioning data transmission delays.

[0031] According to an embodiment of the present invention, the step of issuing a safety warning to travelers at the target scenic spot based on the corrected location data specifically includes: Construct a set of electronic fence area boundary coordinates based on the safe location boundary of the target attraction, and calculate the minimum Euclidean distance between the traveler's location and the electronic fence area boundary coordinates based on the corrected location data; When the minimum Euclidean distance is less than a preset distance threshold, the traveler's movement trajectory data is acquired, and the movement trend is predicted based on the movement trajectory data using a time series analysis algorithm to obtain the traveler's predicted movement trajectory. The predicted time for a traveler to reach the electronic fence boundary area is calculated based on the predicted movement trajectory. When the predicted time is less than the time warning threshold, the traveler is marked as a location risk person, a location warning information of the location risk person is generated and sent to the staff device, and a warning call alarm is issued to the location risk person's mobile terminal. When the minimum Euclidean distance is not less than a preset distance threshold, the traveler is marked as a person in a safe location, and the flow of people in the location analysis area where the person in the safe location is located is monitored in real time. The flow of people is compared with the preset flow of people threshold. When the change in pedestrian traffic exceeds the pedestrian traffic threshold, the performance characteristics of the location data transmission in that area are extracted. The location data transmission delay is predicted by combining the current change in pedestrian traffic data. Based on the predicted location data transmission delay, the frequency of location data collection by the location acquisition platform for people in safe locations is reduced, thus forming a safety early warning strategy for tourists at the target scenic spot.

[0032] It should be noted that, due to potential safety risks associated with travelers approaching the electronic fence area, the system calculates the minimum Euclidean distance between the corrected location data and the electronic fence boundary. When this distance is less than a preset threshold, the system automatically acquires the traveler's movement trajectory and predicts its movement trend, thereby estimating the predicted time of their arrival at the electronic fence area. When the predicted time is lower than the time warning threshold, the system can mark the traveler as a location risk person in real time and generate accurate location warning information to be sent to staff devices. At the same time, it issues a warning call alarm to the traveler's mobile terminal, thereby enabling real-time monitoring and intervention of potentially risky personnel. This effectively prevents people from accidentally entering dangerous areas or causing accidents, improves the initiative and response speed of scenic area safety management, reduces the pressure of manual patrols, and ensures the real-time safety of travelers. Since travelers in safe areas may still experience location data delays or monitoring blind spots due to dense crowds, the system marks travelers with a minimum Euclidean distance of at least a preset threshold as safe location personnel. This allows the system to monitor real-time changes in pedestrian traffic in the location analysis area and compare this data with the preset pedestrian traffic threshold. When pedestrian traffic exceeds the threshold, the system further extracts the location data transmission performance characteristics of the area and predicts potential location data transmission delays based on current pedestrian traffic changes. This dynamically adjusts the location data collection frequency of the location acquisition platform for safe location personnel, ensuring both the continuity and accuracy of monitoring while reducing network pressure and data redundancy in high-density areas.

[0033] According to an embodiment of the present invention, it further includes: Obtain GPS positioning data and inertial sensor data from mobile terminals of tourists within the target scenic spot, and align the inertial sensor data, including acceleration data and angular velocity data, with the GPS positioning data by timestamp to construct a synchronous motion dataset; The velocity and angular velocity change rates of the travelers within a preset time window are calculated based on the synchronous motion dataset and compared with the displacement change rate in the GPS positioning data to obtain a velocity-displacement change difference index. An abnormal motion discrimination vector is constructed based on the velocity-displacement change difference index and the instantaneous acceleration peak value of the inertial sensor. The abnormal motion discrimination vector is used to identify the motion state of the traveler at the current location. When the identification result is an abnormal motion state, the abnormal motion marker and corresponding location data are output. The traveler location prediction model is adjusted based on the abnormal motion markers. Inertial sensor data under abnormal motion conditions is used as high-priority input to correct the model's prediction of the traveler's future location, resulting in corrected predicted location data.

[0034] According to an embodiment of the present invention, the step of identifying abnormal movement states of travelers based on abnormal movement discrimination vectors specifically includes: Extract the acceleration vector change trajectory of the traveler in the synchronous motion dataset over a continuous time period, construct an acceleration feature sequence based on the acceleration vector change trajectory, and calculate the correlation between the acceleration feature sequence and the displacement change sequence in the corresponding time period in the GPS positioning data to obtain the acceleration-displacement coupling feature. A multimodal motion feature vector is constructed based on the acceleration-displacement coupling characteristics and the rate of change of angular velocity; The multimodal motion feature vector is input into the abnormal motion recognition model. The motion state label of the traveler at the current moment is obtained through the preset speed change discrimination rule. When the motion state label is an abnormal motion state, the corresponding abnormal motion position prediction correction parameters are generated. The offset prediction results in the location prediction model are weighted and updated according to the prediction correction parameters to obtain the predicted actual location data under abnormal motion.

[0035] It should be noted that in actual safety warning scenarios for tourists in scenic areas, relying solely on GPS positioning data often suffers from lag and instability, especially when people exhibit abnormal movement behaviors (such as running, falling, or rapid escape). Due to limitations in GPS sampling frequency and signal transmission delay, the location prediction model struggles to capture sudden changes in motion in a timely manner, resulting in tourists not being identified and warned in time before entering dangerous areas, thus creating safety hazards. This paper addresses this issue by introducing inertial sensor data from mobile terminals onto GPS positioning data, aligning acceleration and angular velocity with GPS data through timestamps to construct a synchronous motion dataset. Based on the rate of change of velocity, rate of change of displacement, and peak instantaneous acceleration, a difference index is calculated to construct an abnormal motion discrimination vector, enabling real-time identification of abnormal movement states. When an abnormal movement state is detected, the inertial sensor data is used as a high-priority input to dynamically adjust the location prediction model, obtaining corrected predicted location data, ensuring that location prediction can quickly respond to sudden changes in people's states. Furthermore, by extracting acceleration change trajectories, constructing acceleration feature sequences, and performing correlation analysis with GPS displacement change sequences, a multimodal motion feature vector is generated by combining the angular velocity change rate. This vector is then input into the abnormal motion recognition model, and abnormal motion correction parameters are generated through velocity change discrimination rules. The location prediction model is then updated with weights. This not only compensates for the delay and bias issues of GPS positioning but also enables rapid identification and location prediction correction at the first sign of abnormal behavior, significantly improving the real-time performance and accuracy of safety warnings for travelers and reducing the risk of people entering dangerous areas.

[0036] Figure 4 A block diagram of a traveler safety early warning system based on real-time location data management according to the present invention is shown.

[0037] A second aspect of the present invention also provides a traveler safety early warning system based on real-time location data management. The system includes a memory 401, a processor 402, and a communication interface 403. The memory includes a traveler safety early warning method program based on real-time location data management. The communication interface is used for data connection and communication between the memory and the processor. When the processor executes the traveler safety early warning method program based on real-time location data management, it performs the following steps: Construct a location data collection platform for tourists at the target scenic spot, acquire GPS positioning data and actual location data of tourists within the target scenic spot from the location data collection platform, determine the positioning accuracy characteristics and positioning data transmission performance characteristics of different locations at the target scenic spot, and construct a positioning accuracy feature map and a positioning data transmission performance feature map; The positioning accuracy feature map and the positioning data transmission performance feature map are fused based on Gaussian process regression to construct a joint feature vector. The positioning data acquisition accuracy offset value of the location acquisition platform at different locations is determined based on the joint feature vector, including the offset direction and offset distance value. Based on the decision tree model and the location acquisition platform's accuracy offset values ​​for different locations, a personnel location offset prediction model is constructed. Based on the personnel location offset prediction model, the location of travelers on the location acquisition platform is predicted, and the predicted actual location data is output. The traveler's location on the location acquisition platform is corrected based on the predicted actual location data to obtain corrected location data; Safety warnings are issued to travelers at the target attractions based on the corrected location data.

[0038] This invention discloses a method and system for traveler safety early warning based on real-time location data management. The method constructs a traveler location acquisition platform to obtain GPS positioning data and actual location data of travelers, forming a positioning accuracy feature map and a data transmission performance feature map. Gaussian process regression is used to fuse the maps, generating a joint feature vector to determine the acquisition accuracy offset value at different locations. Furthermore, a personnel location offset prediction model is established using a decision tree model to predict the actual location of travelers. Based on the prediction results, the positioning data of the acquisition platform is corrected to obtain higher accuracy location data. Finally, real-time safety early warnings are issued to travelers based on the corrected data. This method can effectively overcome the errors caused by fluctuations in positioning accuracy and unstable transmission performance in complex scenic environments, improving the accuracy and reliability of traveler positioning.

[0039] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple units or components can be combined, or integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the various components shown or discussed can be through some interfaces, and the indirect coupling or communication connection between devices or units can be electrical, mechanical, or other forms.

[0040] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units. They may be located in one place or distributed across multiple network units. Some or all of the units may be selected to achieve the purpose of this embodiment according to actual needs.

[0041] In addition, in the various embodiments of the present invention, each functional unit can be integrated into one processing unit, or each unit can be a separate unit, or two or more units can be integrated into one unit; the integrated unit can be implemented in hardware or in the form of hardware plus software functional units.

[0042] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0043] Alternatively, if the integrated units of this invention are implemented as software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of this invention, or the parts that contribute to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROM, RAM, magnetic disks, or optical disks.

[0044] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for traveler safety early warning based on real-time location data management, characterized in that, Includes the following steps: Construct a location data collection platform for tourists at the target scenic spot, acquire GPS positioning data and actual location data of tourists within the target scenic spot from the location data collection platform, determine the positioning accuracy characteristics and positioning data transmission performance characteristics of different locations at the target scenic spot, and construct a positioning accuracy feature map and a positioning data transmission performance feature map; Specifically, the positioning accuracy characteristics and positioning data transmission performance characteristics are obtained by the platform using a kernel density estimation algorithm to analyze the distribution of actual location data, dividing the scenic spot into multiple positioning analysis areas, and pairing GPS data with actual location data at the same timestamp using a spatiotemporal alignment method to calculate the positioning accuracy characteristics of each area; at the same time, the dynamic changes in transmission delay are obtained by combining the time difference between GPS data generation and reception in different time periods, and the correlation between this and the flow of people is analyzed to establish a correlation model between flow of people and transmission delay, thereby obtaining the positioning data transmission performance characteristics under different flow of people density; The positioning accuracy feature map and the positioning data transmission performance feature map are fused based on Gaussian process regression to construct a joint feature vector. The positioning data acquisition accuracy offset value of the location acquisition platform at different locations is determined based on the joint feature vector, including the offset direction and offset distance value. Based on the decision tree model and the location acquisition platform's accuracy offset values ​​for different locations, a personnel location offset prediction model is constructed. Based on the personnel location offset prediction model, the location of travelers on the location acquisition platform is predicted, and the predicted actual location data is output. The traveler's location on the location acquisition platform is corrected based on the predicted actual location data to obtain corrected location data; Safety warnings are issued to travelers at the target attractions based on the corrected location data.

2. The traveler safety early warning method based on real-time location data management according to claim 1, characterized in that, The aforementioned tourist location collection platform for the target scenic spot acquires GPS positioning data and actual location data of tourists within the target scenic spot from the location collection platform, determines the positioning accuracy characteristics and positioning data transmission performance characteristics of different locations within the target scenic spot, and constructs a positioning accuracy feature map and a positioning data transmission performance feature map, specifically as follows: A location data collection platform for tourists at the target scenic spot is constructed. The location data collection permissions of the mobile terminals of tourists entering the target scenic spot are obtained. The location data collection permissions are connected to the location data collection platform to collect the GPS positioning data of tourists in real time. At the same time, the actual location data of tourists are obtained based on the pre-installed visual positioning system in the target scenic spot. The actual location data distribution density of each area within the target scenic spot is calculated based on the kernel density estimation algorithm within a first preset time period. The target scenic spot is then divided into several location analysis areas based on the actual location data distribution density. Perform spatiotemporal alignment of GPS positioning data and actual location data at the same timestamp to obtain a set of location data pairs. Within each positioning analysis area, calculate the mean Euclidean distance between GPS positioning data and actual location data in each location data pair as the positioning accuracy feature of that area. The generation timestamp of GPS positioning data and the reception timestamp of GPS positioning data of the location acquisition platform are obtained for each positioning analysis area within a second preset time period. Based on the generation timestamp and the reception timestamp, the data transmission delay variation data within the second preset time period is determined. Based on the real-time location data, determine the pedestrian flow change data of each location analysis area within a second preset time period, perform correlation analysis between the pedestrian flow change data and the data transmission delay change data, establish a pedestrian flow-transmission delay correlation model, and obtain the location data transmission performance characteristics of each location analysis area under different pedestrian flows. Based on the Kriging interpolation method, the positioning accuracy features and positioning data transmission performance features are interpolated within the positioning analysis area of ​​the target scenic spot, and mapped to the position coordinates of each positioning analysis area of ​​the target scenic spot to generate the positioning accuracy feature map and the positioning data transmission performance feature map of the target scenic spot.

3. The traveler safety early warning method based on real-time location data management according to claim 2, characterized in that, The method involves fusing the positioning accuracy feature map and the positioning data transmission performance feature map based on Gaussian process regression to construct a joint feature vector. Based on this joint feature vector, the positioning data acquisition accuracy offset value of the location acquisition platform at different locations is determined, including the offset direction and offset distance. Specifically: A Gaussian process regression algorithm is introduced, and the position coordinates of each positioning analysis region in the positioning accuracy feature map and the positioning data transmission performance feature map are used as the input space of the Gaussian process regression. Based on the positioning accuracy feature map and the positioning data transmission performance feature map, the positioning accuracy feature value and the positioning data transmission performance feature value of each positioning analysis region are extracted as observation values ​​to construct a training dataset. The covariance function of Gaussian process regression is initialized as a radial basis function. Initial hyperparameters, including length scale and signal variance, are set. The hyperparameters of Gaussian process regression are optimized by the maximum likelihood estimation method to obtain the optimized covariance function. The joint covariance matrix between the positioning accuracy feature value and the positioning data transmission performance feature value is calculated based on the optimized covariance function. The Cholesky decomposition of the joint covariance matrix is ​​used to obtain the lower triangular matrix. The weight coefficients of the Gaussian process regression are then solved based on the lower triangular matrix. Input the location coordinates of each location analysis area into the trained Gaussian process regression model, and output the fused prediction value of the location accuracy feature and the location data transmission performance feature. Construct a joint feature vector for each location analysis area based on the fused prediction value. The feature gradient change between adjacent positioning analysis areas is calculated based on the joint feature vector. The positioning data acquisition accuracy offset direction is solved by the gradient descent method. The gradient change rate of the joint feature vector in the offset direction is calculated. The gradient change rate is multiplied by the preset offset distance conversion coefficient to obtain the offset distance value. The positioning data acquisition accuracy offset value is constructed by the positioning data acquisition accuracy offset direction and the offset distance value.

4. The traveler safety early warning method based on real-time location data management according to claim 2, characterized in that, The process involves constructing a personnel location offset prediction model based on a decision tree model and the location acquisition platform's data acquisition accuracy offset values ​​for different locations. This model is then used to predict the location of travelers on the location acquisition platform, outputting predicted actual location data. Specifically: The location acquisition platform collects positioning data at different locations according to a preset data volume ratio to construct the training data matrix and validation set data of the decision tree model. The positioning data collection accuracy offset value of each positioning analysis area is used as the input feature, and the Euclidean distance between the GPS positioning data and the actual location data of the corresponding area is used as the output label. The Gini coefficient is used as the splitting criterion for decision tree nodes. The training data matrix is ​​recursively divided into features. At each node, the feature of the positioning data acquisition accuracy offset value that makes the Gini coefficient decrease the most is selected for splitting until the preset tree depth or node sample number threshold is reached, and the initial decision tree model is generated. The initial decision tree model is optimized by post-pruning algorithm. The change in validation set error after each non-leaf node is replaced with a leaf node is calculated. When the error change is less than a preset threshold, pruning operation is performed to obtain the personnel position offset prediction model. The system acquires real-time GPS positioning data of travelers collected by the location acquisition platform, extracts the positioning data acquisition accuracy offset value feature of the positioning analysis area where the GPS positioning data is located, inputs the positioning data acquisition accuracy offset value feature into the trained personnel location offset prediction model, matches the feature value layer by layer through the splitting rules of the decision tree, and finally reaches the leaf node to obtain the corresponding predicted offset distance and predicted offset direction. The predicted actual location data of travelers is obtained by vector superposition calculation based on the coordinate position of real-time GPS positioning data, the predicted offset distance, and the predicted offset direction.

5. A traveler safety early warning method based on real-time location data management according to claim 1, characterized in that, The step of correcting the traveler's location on the location acquisition platform based on the predicted actual location data to obtain corrected location data specifically involves: The predicted actual location data is compared with the GPS location data of travelers collected in the location acquisition platform to determine the deviation between the GPS location data collected by the location acquisition platform and the predicted actual location data. If the deviation value is greater than the preset value, the GPS location data collected by the location acquisition platform is replaced according to the predicted actual location data to obtain the corrected location data of the traveler in the location acquisition platform.

6. A traveler safety early warning method based on real-time location data management according to claim 2, characterized in that, The step of issuing safety warnings to travelers at the target scenic spot based on the corrected location data specifically includes: Construct a set of electronic fence area boundary coordinates based on the safe location boundary of the target attraction, and calculate the minimum Euclidean distance between the traveler's location and the electronic fence area boundary coordinates based on the corrected location data; When the minimum Euclidean distance is less than a preset distance threshold, the traveler's movement trajectory data is acquired, and the movement trend is predicted based on the movement trajectory data using a time series analysis algorithm to obtain the traveler's predicted movement trajectory. The predicted time for a traveler to reach the electronic fence boundary area is calculated based on the predicted movement trajectory. When the predicted time is less than the time warning threshold, the traveler is marked as a location risk person, a location warning information of the location risk person is generated and sent to the staff device, and a warning call alarm is issued to the location risk person's mobile terminal. When the minimum Euclidean distance is not less than a preset distance threshold, the traveler is marked as a person in a safe location, and the flow of people in the location analysis area where the person in the safe location is located is monitored in real time. The flow of people is compared with the preset flow of people threshold. When the change in pedestrian traffic exceeds the pedestrian traffic threshold, the performance characteristics of the location data transmission in that area are extracted. The location data transmission delay is predicted by combining the current change in pedestrian traffic data. Based on the predicted location data transmission delay, the frequency of location data collection by the location acquisition platform for people in safe locations is reduced, thus forming a safety early warning strategy for tourists at the target scenic spot.

7. A traveler safety early warning system based on real-time location data management, characterized in that, The traveler safety early warning system based on real-time location data management includes a storage unit and a processor. The storage unit includes a traveler safety early warning method program based on real-time location data management. When the processor executes the traveler safety early warning method program based on real-time location data management, it performs the following steps: Construct a location data collection platform for tourists at the target scenic spot, acquire GPS positioning data and actual location data of tourists within the target scenic spot from the location data collection platform, determine the positioning accuracy characteristics and positioning data transmission performance characteristics of different locations at the target scenic spot, and construct a positioning accuracy feature map and a positioning data transmission performance feature map; Specifically, the positioning accuracy characteristics and positioning data transmission performance characteristics are obtained by the platform using a kernel density estimation algorithm to analyze the distribution of actual location data, dividing the scenic spot into multiple positioning analysis areas, and pairing GPS data with actual location data at the same timestamp using a spatiotemporal alignment method to calculate the positioning accuracy characteristics of each area; at the same time, the dynamic changes in transmission delay are obtained by combining the time difference between GPS data generation and reception in different time periods, and the correlation between this and the flow of people is analyzed to establish a correlation model between flow of people and transmission delay, thereby obtaining the positioning data transmission performance characteristics under different flow of people density; The positioning accuracy feature map and the positioning data transmission performance feature map are fused based on Gaussian process regression to construct a joint feature vector. The positioning data acquisition accuracy offset value of the location acquisition platform at different locations is determined based on the joint feature vector, including the offset direction and offset distance value. Based on the decision tree model and the location acquisition platform's accuracy offset values ​​for different locations, a personnel location offset prediction model is constructed. Based on the personnel location offset prediction model, the location of travelers on the location acquisition platform is predicted, and the predicted actual location data is output. The traveler's location on the location acquisition platform is corrected based on the predicted actual location data to obtain corrected location data; Safety warnings are issued to travelers at the target attractions based on the corrected location data.

Images